A Novel Optical Control Plane for Switching an Electro-Optical Hybrid Node in Translucent WDM Optical Networks

Home , Optical filter

Communications and Network, 2013, 5, 57-68 http://dx.doi.org/10.4236/cn.2013.51006 Published Online February 2013 (http://www.scirp.org/journal/cn)

Sridhar Iyer1, Shree Prakash Singh2 1Department of ECE, NIIT University, Rajasthan, India 2Division of ECE, Netaji Subhas Institute of Technology, New Delhi, India Email: [email protected], [email protected], [email protected]

Received December 11, 2012; revised January 12, 2013; accepted February 13, 2013

ABSTRACT In a translucent network scenario, development of an optical control plane (OCP) that is aware of the location and number of available regenerators and all-optical wavelength converters (AOWCs) is of paramount importance. How- ever, current generalized multiprotocol label switching (GMPLS) protocol suite does not consider the distribution of regenerator and AOWC availability information to all the network nodes. In this paper, we propose a novel optical control plane (OCP) architecture that 1) disseminates information about network components (i.e. regenerators and AOWCs) to all the network nodes; and 2) evaluates candidate routes which use fewest amounts of network components. Performance of the proposed OCP is compared with a recently proposed hybrid OCP approach in terms of blocking performance, number of deployed components and lightpath establishment setup times. The obtained simulation results show that the proposed OCP approach demonstrates low connection blocking and establishes lightpaths by 1) minimizing the overall network cost owing to the deployment of minimum total number of network components; and 2) demonstrating acceptable lightpath establishment setup times at all traffic loads. Further, the proposed OCP methodology is compatible and suitable for controlling the operations of a novel electro-optical hybrid translucent node which is a latency efficient technology capable of delivering a cost effective implementation suitable for large scale deployment.

Keywords: Optical Control Plane; Translucent WDM Optical Network; Physical Layer Impairments; Regenerators; All-Optical Wave-Length Converters; GMPLS; PLI-SQARWA

1. Introduction due to signal quality or wavelength contention requirements, the usual solution is to regenerate the signal(s) at In opaque wavelength division multiplexed (WDM) op- intermediate nodes. The imminent transparency disrup- tical networks, signals undergo optical-electrical-optical (OEO) conversions at each intermediate node thus pur- tion of signals due to OEO conversions and increase in veying a-priori admissible quality of transmission (QoT) the overall network cost by the introduction of regenera- at the expense of costly 3R (re-amplify, reshape and tors has led to the emergence of translucent WDM net- re-time) electrical regenerators which contribute to both, works which exploit sparse placement (i.e., installed in a capital and operational expenditures (CapEx and OpEx). limited number of nodes) of regenerators, thus repre- With the new generation of commercially available opti- senting a promising trade-off between transparent and cal network elements, the paradigm shift towards trans- opaque networks [2]. Hence, translucent networks are parent (all-optical networks) WDM optical networks has able overcome the reach limitations due to QoT degrada- evolved wherein; signals are transmitted in the optical tion by PLIs, simultaneously keeping network cost lim- domain from source to the destination. The flexibility ited [3]. and high capacity provided by transparent networks In translucent WDM networks, for each lightpath re- though are faced with physical layer impairment (PLIs) quest, a route must be found, specifying both, link re- inherent in the fiber and further, these networks suffer sources (i.e., wavelength channels) and node resources from inefficient wavelength utilization as a connection (i.e., OEO regenerators and all-optical wavelength con- request may be rejected due to unavailability of a com- verters-AOWCs) to be used. In addition, a node must mon wavelength on all the links along the chosen route also be provided with the information pertaining to 1) [1]. For the above reasons, in order to provision demands link and node QoT parameters and wavelength conver-

Copyright © 2013 SciRes. CN 58 S. IYER, S. P. SINGH sion (WC) requirements; and 2) availability of installed latter case, the hybrid node uses a regenerator for both, regenerators and AOWCs at the nodes. Owing to the regeneration and WC. However, if only wavelength con- former information, a node is aware of the QoT degrada- tention is to be resolved, the hybrid node resorts to an tion and WC requirements along every eligible path AOWC instead of a regenerator. whereas, owing to the latter information, in the case of In a translucent WDM network scenario, the imple- critical QoT and/or wavelength collisions, a node can mentation of an efficient optical control plane (OCP) that choose the regenerators and/or AOWCs to be exploited. provisions distributed signaling and routing mechanisms Existing studies on translucent network design focus on throughout the network is of paramount importance. The 1) physical layer impairment-routing and wavelength GMPLS based OCP, which is the de facto standard and assignment (PLI-RWA) approaches for incorporating implementation of the OCP used for disseminating in- PLIs during lightpath computation; 2) selecting regen- formation in translucent WDM networks, is based on eration sites and number of regenerators to be deployed two main distributed protocols: 1) the Open Shortest Path on these sites (regenerator placement (RP) problem) First with Traffic Engineering Extensions (OSPF-TE) and/or; 3) given sparse placement of regenerators, se- routing protocol that floods the network topology and TE lecting which of these regenerators to use (regenerator information used for path computation; and 2) the Re- allocation (RA) problem) [4]. These existing studies pre- source Reservation Protocol with Traffic Engineering sent the following two major drawbacks: Extensions (RSVP-TE) signaling protocol that is respon- 1) They are based on shortest path (SP) routing proto- sible for the actual lightpath establishment. Figure 1 cols since; SP routing sufficiently achieves the desired shows the existing optical network layers wherein; the performance. However, the SP approach contributes sig- OCP is responsible for TE, route computation and significantly to network congestion and also comes short of naling; while the data plane is responsible for switching provisioning efﬁcient network utilization, since trafﬁc the data belonging to different connections based on characteristics and resource availability are not a part of OCP instructions. its routing decisions. Further, owing to the existence of However, standard GMPLS does not account for the non-uniform PLIs and heterogeneity of real optical net- QoT, regenerator and AOWC information and hence, no works, SP may not always demonstrate the best signal solutions have been standardized yet for the issue of en- quality. Hence, it may occur that any other route(s), apart compassing QoT, regenerator or AOWC information from the SP, exhibits better signal quality, thus requiring within the GMPLS-controlled translucent networks [7]. fewer intermediate regenerators, in turn, lowering the The presence of regenerators and AOWCs necessitates overall network cost. the OCP to decide on which regenerators and/or AOWCs 2) They assume that regenerators are also equipped to be used while setting up 1) a translucent lightpath or 2) with WC capability. In cases when signal(s) require only an end-to-end lightpath composed of multiple transpar- WC, such an assumption may not be viable as OEO con- ent sections. In order to solve QoT distribution issues, version leads to time delay. Thus, for only wavelength studies in [8-10] have proposed the inclusion of QoT contention resolution, it is viable to use an AOWC in parameters within both, the RSVP-TE and OSPF-TE order to reduce the delay introduced by OEO conversions. protocols. Recently, studies focussing on GMPLS exten- For cases when only regeneration or simultaneous regen- sions in the context of regenerator information dissemi- eration and WC are required, a regenerator can be used. nation have been proposed [11-13]. In order to tackle the above mentioned drawbacks of existing translucent network design studies, in our previous work, we have proposed: 1) An innovative PLI-Signal Quality Aware RWA (PLI-SQARWA) algorithm [5] that 1) guarantees zero blocking due to signal degradation and wavelength contention, and 2) minimizes the total required network components by a) considering signal quality to route connections over paths requiring fewest numbers of regenerators, and b) maximally using placed regenerators for WC before resorting to AOWCs, and 2) An electro-optical hybrid translucent node architecture [6] that uses regenerators if a) the Quality-factor (Q-factor) at any node falls below the threshold or b) the Q-factor at any node falls below the threshold and si- Figure 1. Optical network layers: control plane and data multaneously there is also wavelength contention. In the plane.

In [11,12], extensions to both, RSVP-TE and OSPF- ing-based OCP and routing-based OCP show that hybrid TE protocols are proposed in order to account for shared OCP exhibits the lowest blocking probability, but at the regenerators in translucent networks. In [11], the authors cost of longer setup delays and increased number of re- have proposed lightweight GMPLS extensions to account generators for each path. The authors conclude that hy- for shared regenerators in translucent networks during brid OCP is able to conquer the limitations of existing lightpath establishment. Firstly, a signaling approach approaches under various traffic conditions, and is hence based on dynamic estimation of optical signal quality a suitable solution for connection management in trans- during the lightpath setup process is applied, which is lucent WDM networks. then extended to encompass the presence of shared re- However, the aforementioned studies do not consider generators. An intermediate node designates itself for distribution of AOWC information to the network nodes. regeneration during the signaling phase if, QoT of a These studies assume that the transparent lightpath sec- transparent section, terminating at the node, falls below tions comprising the translucent lightpath are subject to the pre-defined threshold value. Results of the study wavelength-continuity constraint. In this paper, we con- show that networks demonstrating few critical situations trive an architecture that underpins the OCP design can be setup without 1) significantly affecting the light- problem in translucent networks. The contributions of path blocking probability; and 2) compromising on the this paper are twofold: routing protocol scalability.  A novel OCP architecture is proposed which: 1) en- The authors in [12] have proposed GMPLS extensions sures that a node designated for regeneration and/or to encompass the presence of shared regenerators without WC reserves regenerator and/or AOWC for a given routing protocol modifications. The regenerator object lightpath, and further, also performs regeneration (RO) is introduced within the RSVP-TE to identify the and/or WC, and 2) unlike previous studies [11,12]; in nodes designated for regeneration. In addition, other ex- which regenerator information is utilized only by the tensions to RSVP-TE (i.e., the regenerator availability source node to setup a lightpath; distributes the re- object—RAO) and to OSPF-TE (i.e., the regenerator generator and/or AOWC capability and availability information node state advertisement—RI-NSA) are information to all the nodes by utilizing the regenera- proposed to distribute regenerator availability informa- tor database (RD) and AOWC database (AOWCD). tion. In order to setup a lightpath, the gathered regenera- The RD and AOWCD are installed in each node and tor information is utilized only by the source node and store the number of installed and/or available regen- the results show that low blocking is obtained without erators and AOWCs at each node. Thus, the proposed requiring any routing protocol modification. OCP approach is compatible and suitable for control- The authors in [13] have concentrated on the study of ling the switching operations of the hybrid node. OCP and management approaches for translucent WDM  Deviating from existing studies that use SP routing networks. A hybrid OCP is proposed, which combines protocols, the PLI-SQARWA algorithm is employed the best features of both, routing based information up- for RWA. PLI-SQARWA minimizes connection dating and signaling-based data collection approaches blocking and also results in minimum number of thus, necessitating extensions to both, routing and sig- network components being deployed, thus also mini- naling protocols. When the connection request arrives at mizing the overall network cost. the source node, first a k hybrid weight shortest path It must be noted that, storing the regenerator and first (k-HW-SPF) route computation algorithm is imple- AOWC information at only the source node and further, mented to provide up to k possible routes according to this information being used only by the source node to the information in Simplified Global-Resource Related setup the lightpaths, leads to high congestion and block- Database (SG-RRD) and Simplified Global-Physical Im- ing. However, regenerators and AOWCs information, pairments Database (SG-PID). The calculated routes are when disseminated to all the network nodes, can consid- then passed to RSVP-TE extension module, and paral- erably reduce the congestion and blocking. This can be lel transmitted by k Path (i.e. message denoting the re- explained as follows: for only the source node establish- quest for a connection) messages. In the forward direc- ing the lightpaths by utilizing the regenerator and tion, every traverse node, before propagating the Path AOWC information, in the case of a node failure along message, updates the collected data by adding its own the path, the Path message(s) will be required to traverse dynamic parameters and resources state to the corre- back all the way to the source node in order to retrieve sponding objects of RSVP message. At the same time, the information hence, increasing the lightpath setup the RA strategy is implemented for signal quality estima- attempts and setup time. In case of many such failures, tion and RA. At the destination, node picks the first in- network congestion will result leading to connection coming Path message and queues the other arriving mes- blocking. On the other hand, in the case of regenerator sages. The results comparing hybrid OCP with signal- and AOWC information being disseminated to all the

Copyright © 2013 SciRes. CN 60 S. IYER, S. P. SINGH network nodes, in the event of a node failure along the using standard non-zero dispersion shifted fibers and path, regenerator and AOWC information can be re- erbium-doped fiber amplifiers (EDFAs) are deployed trieved from the neighbouring nodes, hence reducing the every fiber span (typically every 80Km) in order to re- lightpath setup attempts and setup time. This in turn will cover from fiber losses. The placement of hybrid nodes also reduce the network congestion and connection (i.e., required regenerators and AOWCs) in the network blocking. In addition, storing the information at only the is chosen by the PLI-SQARWA algorithm. Figure 3 source node leads to setup of only a few parallel light- shows the architecture of the proposed OCP for translu- paths whereas, by distributing it to all network nodes, the cent WDM network wherein; the switching fabric com- information can be exploited for many parallel lightpath prises of the electro-optical hybrid nodes. setups and also for setup of subsequent lightpath requests As proposed in [14], it is assumed that at each node, [12]. the OCP stores wavelength availability information in the The remainder of the paper is organized as follows. traffic engineering database (TED), which is updated Section 2 describes the translucent network model con- through the signal quality aware routing (SQAR-TE) sidered in the study. In Section 3, we describe details of flooding [5]. Moreover, each node maintains an updated the proposed OCP approach. Section 4 presents the nu- knowledge of the optical layer by storing the QoT parameters and WC requirements in the QoT parameter merical results wherein; we discuss the simulation as- database (QPD) and WC requirement database (WCD) sumptions, the performance metrics used in the study, the respectively. Therefore, each node is able to determine analysis to evaluate setup times for the lightpaths, and the feasibility of a path in terms of QoT and wavelength the simulation results. Finally, section 5 presents the continuity constraint. Further, following the study in [12], conclusion of this study. the regenerator object (RO) and the AOWC object (AOWCO), which contain an entry of each node desig- 2. Network Model nated for regeneration and WC respectively, are also in- The considered translucent network comprises of N cluded within the signaling messages. nodes and L bi-directional links, with each link support- ing W wavelengths per direction. For the fulfillment of 3. Optical Control Plane Design regeneration and/or WC requirement, we resort to the deployment of the hybrid node as shown in Figure 2. 3.1. Problem Definition The wavelength converter within the hybrid node is In this sub-section we describe the OCP design problem based on four wave mixing (FWM) in semiconductor statement which can be stated as follows: optical amplifier (SOA) and is assumed to be ideal or full Given range wavelength converter (FRWC), i.e. it can convert 1) A network topology; any input wavelength to any output wavelength [5,6]. 2) A set of available wavelengths per fiber link; A lightpath transiting such a node may be switched 3) The offered traffic with a set of static demands; transparently (L1) or directed to the regenerator pool (L2) 4) An admissible Q-factor threshold value i.e. Qth. if 1) it requires only regeneration or; 2) it simultaneously Objective requires both regeneration and WC. If only wavelength To evaluate, either a feasible transparent path (if it contention is to be resolved, the lightpath is directed to exists) or a translucent path (comprised of multiple the wavelength converter (L3). Fiber links are deployed transparent sections), with minimum number of network

Figure 2. Electro-optical hybrid translucent node architecture.

Step 3. If a transparent lightpath is available, go to phase 2, else go to step 4. Step 4. If no transparent lightpath is available due to wavelength collision and/or unacceptable QoT, run 1) the wavelength assignment (WA) & wavelength converter placement (WCP) algorithm phase, and 2) RP algorithm phase, of the PLI-SQARWA algorithm, which ensures the minimization of the overall network cost. Go to step 5. Step 5. Designate the hybrid nodes for regeneration and/or WC based on regenerator and AOWC information stored in the RD and AOWCD respectively of every

node. Go to step 6. Figure 3. Architecture of the proposed OCP. Step 6. Establish the translucent lightpath by switch-

ing the deployed hybrid nodes, as per the requirement of components (i.e. regenerators and AOWCs) deployed 1) regeneration, 2) simultaneous regeneration and WC, or along all the links, in effect, minimizing the overall net- 3) only WC. In other words, by appropriately switching work cost, and simultaneously ensuring minimum con- and using the designated hybrid nodes, partition the se- nection blocking and setup time for the lightpath estab- lected candidate lightpath into appropriate number of lishment. transparent sections with acceptable QoT and/or wave- Subject to length continuity along the links, and go to phase 2. Signal quality constraint: For any network lightpath, at its destination node, the corresponding Q-factor value Phase II. Signaling Phase must not fall below the pre-defined threshold value. In this phase, both, the transparent and translucent Minimum cost constraint: For any network lightpath, lightpaths resort to a signaling session for reserving the regeneration and/or WC requirements must be satisfied required components in order to establish the lightpaths. ensuring the deployment of minimum number of com- The signaling session comprises of the following ten ponents along all the links. steps: Setup Delay Constraint: For any network lightpath, Step 1. At an intermediate node i, upon receipt of the delay between its arrival and establishment must be signaling Path message, check if node i is listed in the minimized and must not exceed a pre-defined setup time RO and/or the AOWCO. If node i is not listed in the RO value. In the current study, the setup time incurred must and/or the AOWCO, go to step2. Else, if node i is listed not exceed a given setup timeout T , which is as- setup in the RO and/or the AOWCO, go to step 3. sumed to be user adjustable. Step 2. Forward the signaling Path message to the next node along the path and designate the next node as 3.2. Proposed Optical Control Plane Approach node i. If this next node (i.e. node i) is an intermediate In this sub-section, we present details of the proposed node, then go to step 1. Else, if this next node (i.e. node i) OCP approach which comprises of two phases. The flow is the destination node, then go to step 6. chart of the proposed OCP is described by the diagram Step 3. Probe the QPD and/or WCD to check the QoT on Figure 4. and/or WC requirement of the transparent section terminating at node i, and go to step 4. Phase I. Routing and Component Deployment Phase Step 4. If the transparent section terminating at node i In this phase, the proposed OCP uses the PLI-SQARWA has an acceptable QoT and/or satisfies the wavelength algorithm in order to 1) select a candidate route based on continuity constraint, from node i, forward the Path signal quality and 2) deploy minimum numbers of net- message without modifying the RO and/or AOWCO, and work components along the selected lightpath. This phase go to step 6. Else, go to step 5. comprises of the following six steps: Step 5. If the transparent section terminating at node i Step 1. For a lightpath request from source node to the does not have an acceptable QoT and/or does not satisfy destination node, compute a candidate path using the the wavelength continuity constraint, reserve the regen- SQAR algorithm, that selects the route based on signal erator and/or AOWC as per the requirement, and further, quality, and go to step 2. modify the RO and/or AOWCO accordingly. In other Step 2. Check the QoT and WC requirements of the words, switch the hybrid node as per the requirement of 1) candidate lightpath by probing the QPD and WCD re- regeneration; 2) simultaneous regeneration and WC, or 3) spectively stored at every node, and go to step 3. only WC, and upgrade both, the RO and/or AOWCO as

Figure 4. Synopsis of the proposed OCP.

Copyright © 2013 SciRes. CN S. IYER, S. P. SINGH 63 required. Go to step 6. Intensity Modulation/Direct Detection (IM/DD) system Step 6. Designate the next node along the lightpath as based on on-off Keying (OOK) and non-return-to-zero node i. If this next node (i.e. node i) is an intermediate (NRZ) modulation [15]. In view of longer optical reach of node, go to step 1. Else, if this next node (i.e. node i) is the signals with the aim of utilizing minimum amount of the destination node, go to Step 7. OEO regenerators, we have considered channels operat- Step 7. At the destination node, upon receipt of the ing at a bit-rate of 10 Gb/s. Further, lightpath QoT eval- Path message, verify the QoT and/or wavelength conti- uation is based on the realistic estimation of signal quality nuity of the last transparent section, and go to step 8. considering the simultaneous impact of three effects viz. Step 8. If the lightpath is not admissible, reserve the stimulated Raman scattering (SRS), FWM, and amplified regenerator and/or AOWC as per the requirement i.e., spontaneous emission (ASE) noise. switch the hybrid node as per the requirement of 1) re- We consider permanent lightpath demands (PLDs) generation; 2) simultaneous regeneration and WC; or 3) which are offline requests that consist of pre-known only WC, and go to step 10. Else go to Step 9. connection demands with data rate equal to full capacity Step 9. If the lightpath is admissible, send a Resv mes- of the wavelength channel and are thus established sage (i.e. message denoting the reservation of required through a full lightpath. In particular, both, the in- number of components for lightpath establishment) car- ter-arrival and the holding times are exponentially dis- rying a copy of the received RO and/or AOWCO, from tributed with a mean time of 1  (with   2 ms) and the destination node, towards all the network nodes along 1 µ (with µ  1 ms) respectively. Let mt  and the lightpath, and go to step 10. It must be noted that the  t denote the mean time and the average service time, Resv message stored at all the nodes along the lightpath, respectively. Hence, the duration of each request is ex- aids in future lightpath establishments, since the network ponentially distributed with an average time that fluctu- is aware of the reserved components. ates between the 1) sum of mean time and average ser- Step 10. Terminate the algorithm. vice i.e.  mt     t  , and 2) mean time mt  . It must be noted that the type of requisite service decides 4. Numerical Results the service time for a demand, which is related to the In this section, we first precise our simulation environ- processing delay, that in turn depends on if the connec- ment and characteristics, then present the performance tion requires 1) only regeneration; or 2) simultaneous metrics used in the study and finally, discuss the obtained regeneration and WC; or 3) only WC. In the current study, numerical results. for WC, since no OEO conversions are required, the processing delay is assumed to be zero whereas, regen- 4.1. Simulation Assumptions eration, which necessitates OEO conversion, is considered to incur a considerable (fixed) amount of processing We conducted extensive simulations on a dual core 2.20 delay. Table 1 summarizes the various transmission sys- GHz computer with 1 GB RAM using the MatLab and tem parameters adopted in our simulations. C++ software in order to investigate the N = 18 node NSFNET network depicted in Figure 5. The network 4.2. Performance Metrics consists of L = 29 bi-directional links with W = 16 wavelengths per fiber link. In the current study, Q-factor In the current study, the following metrics have been threshold takes the value of 6 which corresponds to a bit used for the performance evaluation: error rate (BER) of 10−9. The optical physical layer is 1) Blocking Probability (BP): The ratio between num- modeled as in our previous work, which considers an bers of blocked (rejected) lightpaths and the total number

Figure 5. 18 node NSFNET network.

Table 1. Transmission system parameters. lightpath,  BER the time to run BER estimation and n denotes the number of trials before which a “good” (BER Parameters Values less than threshold) lightpath is calculated or below Wavelength spacing (GHz) 50 which the check for all candidate lightpaths is finished. SMF losses (dB/km) 0.2 On the other hand, for a lightpath, if within the timeout threshold and requiring OEO conversion (i.e. electrical Laser power/channel (dBm/channel) 3 regenerators), the processing time is given as SMF dispersion (ps/nm.km) 17 n Bit Rate (Gbps) 10 Dcprocessin g   LP BER  OEO (2) i1 Switch crosstalk ratio (dB) [16] 32 where  OEO denotes the delay in OEO conversion. Let ASE Factor (nsp) [17] 1.5 the propagation time be denoted by Dcpropagation   , Optical filter bandwidth 10 GHz which is the time involved in carrying the lightpath along Electrical bandwidth of receiver 1 GHz the link between two nodes of the network. Thus, the total set up time for a lightpath “c” is given as Input EDFA gain (dB) 22

Output EDFA gain (dB) 16 DcDcscprocessin g propagation  ,    Converter efficiency (dB) [17] −6.5 Sctotal   if within Tsetup  (3)

−8   ASE power spectral density (W/Hz) 1.63 × 10 Tsetup ,else  (ms) LP 10 Other delays apart from processing and propagation

 BER (ms) 20 delays have been ignored in the analysis and since, AOWC does not involve electrical regeneration, it is as-  OEO (ms) [18] 1 sumed that WC does not introduce any delay. Dcpropagation   (ms) [19] 0.5 ms/100km of fiber

Tsetup (ms) [20] 100 4.4. Simulation Results This sub-section analyses performance of the proposed of lightpath requests. OCP, which is named as P-OCP, by means of compari- 2) Total Number of components: The number of com- son with the recently proposed hybrid-OCP approach, ponents (i.e., regenerators + AOWCs) deployed within which is named as H-OCP [13]. The H-OCP approach the network to establish the lightpath connections. has been executed for range of different values of a set of 3) Average setup time: The average time between the specific parameters that adjust in order to determine how lightpath request arrival and lightpath establishment, the solution space is explored heuristically. The results among all the accepted lightpath requests. shown in this paper correspond to those parameters which provided the best performances and their values 4.3. Setup Time Analysis are reproduced so as to allow the results to be repeatable: For the evaluation of the lightpath setup time, we use the 1) for the computation of k-HW-SPF routes associated to following analysis: A connection demand i.e., a lightpath each demand, the value of k is set to 5; 2) the trace-for- “c” is assumed to incur a total set up time, denoted as ward strategy is used for signaling based RA; 3) the

Sctotal   . The lightpath is blocked if the set up incurs a lightpath quality estimation (LQE) module is used to time that exceeds a given timeout, which is assumed to model the PLI effects, 4) the OCP topology is assumed to be user adjustable with delay bound denoted as Tsetup . be same as that of the data plane, and 5) random selec- Further, as per the requirement(s) of regeneration and/or tion strategy is used for WA. The simulations cover 6 WC, the reservation of the components (i.e. regenerators traffic loads that range from 100 to 500 connection de- and AOWCs), along the lightpath incurs a signaling time, mands and for each load, 10 static traffic matrices are denoted as s c . generated stochastically according to uniform distribu- For a lightpath, if within the timeout threshold and tion. Thus, presented results are average values of ten without requiring OEO conversion (i.e. no electrical re- simulation runs. generators), processing time, Dcprocessin g  is given as In Figure 6, variation of BP with traffic load has been n plotted for both the OCP approaches. It can be observed Dcprocessin g LP BER (1) from the figure that i1 1) For low traffic loads, both, H-OCP and P-OCP where  LP is the time required to find a candidate show zero BP. This can be attributed to the fact that for

Blocking Probability vs Traffic Load 0.012 H-OCP P-OCP 0.01

0.008

0.006

Blocking Probability 0.004

0.002

0 0 50 100 150 200 250 300 350 400 450 500 Traffic Load (PLDs) Figure 6. H-OCP versus P-OCP: blocking probability versus traffic load. both the algorithms: regeneration or 2) simultaneous regeneration and WC a) Regeneration and WC is provisioned which eradi- whereas, for only WC, AOWCs are used. It can be ob- cates blocking due to signal degradation and wavelength served from the figure that: contention respectively, and 1) For low traffic loads, both approaches require ap- b) Owing to low loads, the network is not overloaded proximately the same number of components. and ample resources are available. 2) For moderate to high values of traffic load, H-OCP 2) For intermediate and high traffic loads, blocking requires larger number of components compared to the starts to occur for both the algorithms owing to unavail- proposed P-OCP approach. This occurs due to the fact ability of resources. It can be observed that H-OCP shows that for k-H-OCP (k > 1), large number of regenerators higher blocking compared to the proposed P-OCP ap- are required for each path in order to overcome both, proach. This can be attributed to the fact that: wavelength contention and signal degradation. Thus, a) For H-OCP, as load increases, owing to the use of acceptable blocking performance shown by H-OCP be- K-HW-SPF routing, which is founded on SP routing pro- tween moderate and high loads (as shown in Figure 6) is tocol, higher network overloading occurs, resulting in at the expense of increased number of required compo- unavailability of resources, nents. The increase in the number of total components is b) For P-OCP, rather than the SPs, majority of the by a factor of: “Best Paths” are selected as the candidate routes which a) 1.96 when the load is 400 and, leads to larger resources being available even at higher b) 2.17 when the load is 500. traffic loads. The low blocking observed at higher loads Figure 8 shows the average setup time of the two OCP in the case of P-OCP is due to the fact that PLI- schemes with respect to the traffic load. It can be ob- SQARWA selects few SP routes as candidate routes. served from the figure that It can also be deduced from Figure 6 that for low At low traffic loads (between 50 and 150 demands); loads, both OCP approaches accept all the demands owing to the use of destination-initiated resource reser- whereas, between moderate and high loads, both the ap- vation, the setup delay for H-OCP is more than that of proaches are able to maintain acceptable blocking. P-OCP which is based on source-initiated resource res- In Figure 7, the average total numbers of required ervation. Further, the use of regenerators for both, regen- components, with respect to traffic load, has been plotted eration and WC results in larger delay in the case of for both the OCP approaches. H-OCP compared to P-OCP which uses regenerators for It must be noted that in the H-OCP approach, regen- 1) only regeneration or 2) simultaneous regeneration and erators are used for both, regeneration and WC whereas, WC; and AOWCs for only wavelength contention resolu- in the P-OCP approach, regenerators are used for 1) only tions.

Average Total Number of Components vs Traffic Load 100 Total number of components in H-OCP 90 Total number of components in P-OCP Number of regenerators in P-OCP 80 Number of AOWCs in P-OCP

20 Average Total Number of Components

0 0 50 100 150 200 250 300 350 400 450 500 Traffic Load (PLDs) Figure 7. H-OCP versus P-OCP: number of components versus traffic load.

Average Setup Time vs Traffic Load 90 H-OCP P-OCP 80

50 Average Setup Time (ms)

30 50 100 150 200 250 300 350 400 450 500 Traffic Load (PLDs) Figure 8. H-OCP versus P-OCP: average setup time versus traffic load.

At moderate traffic loads (between 200 and 300 decrease slightly with increase in the load. This is due to mands), the setup delay for H-OCP increases due to ex- the fact that, for the P-OCP approach, information re- cessive time required for multiple lightpath setup-at- quired for lightpath establishment (i.e., number of in- tempts; since k-H-OCP has at most k path setup-attempts stalled and/or available regenerators and/or AOWCs) is according to the path probe information at the destination stored at every network node, which results in lesser [13]. On the contrary, for P-OCP, the setup times in- lightpath setup attempts resulting in lower lightpath setup

Copyright © 2013 SciRes. CN S. IYER, S. P. SINGH 67 times. sidered in Regenerator Placement and Operation of At high traffic loads (beyond 350 demands); the setup Translucent Optical Networks,” Proceedings of IEEE In- delay for H-OCP starts to decrease slightly. This occurs ternational Conference on Transparent Optical Networks (ICTON), Munich, 27 June-1 July 2012, pp. 1-4. since for longer lightpaths, the BP is high, which reduces doi:10.1109/ICTON.2010.5549297 the average setup delay. On the other hand, for P-OCP, [2] J. Berthold, A. Saleh, L. Blair and J. Simmons, “Optical the average setup time increases with network load be- networking: Past, Present, and Future,” IEEE Journal of cause an increasing number of lightpaths are established Lightwave Technology, Vol. 26, No. 9, 2008, pp. 1104- (i.e. BP is very low) by discovering the available com- 1118. doi:10.1109/JLT.2008.923609 ponents within the network. [3] G. Shen and R. S. Tucker, “Translucent Optical Networks: Hence, it can be inferred from Figure 8 that, there ex- The Way Forward,” IEEE Communications Magazine, ists a definitive trade-off between BP and lightpath setup Vol. 45, No. 2, 2007, pp. 48-54. time. Overall, compared to H-OCP, P-OCP is able to doi:10.1109/MCOM.2007.313394 maintain lower BP and lightpath setup times for all traf- [4] S. P. Singh, S. Iyer, S. Kar and V. K. Jain, “Study on fic load values. Mitigation of Transmission Impairments and Issues and Challenges with PLIA-RWA in Optical WDM Net- 5. Conclusions works,” Journal of Optical Communications, De-Gruyter, Vol. 33, No. 2, 2012, pp. 83-101. In this paper, we proposed a novel OCP architecture that doi:10.1515/joc-2012-0015 1) disseminates information about both, regenerators and [5] S. Iyer and S. P. Singh, “A Novel Offline PLI-RWA and AOWCs to all the network nodes; 2) reduces the overall Hybrid Node Architecture for Zero Blocking and Time network cost and connection blocking; and 3) demon- Delay Reduction in Translucent Optical WDM Net- works,” Communications and Networks, Scientific Re- strates acceptable lightpath setup times at all traffic loads. search, Vol. 4, No. 4, 2012, pp. 306-321. In view of a translucent WDM network, performance of doi:10.4236/cn.2012.44036 the proposed OCP (P-OCP) is compared with a recently [6] S. Iyer and S. P. Singh, “A Novel Hybrid Node Architec- proposed hybrid OCP (H-OCP) approach. ture for Reducing Time Delay in Wavelength Division Simulation results show that compared to H-OCP, Multiplexed Translucent Network,” Proceedings of IEEE owing to the use of PLI-SQARWA for RWA, P-OCP is National Conference on Communication (NCC), Kharag- able to 1) maintain acceptable blocking at moderate and pur, 3-5 February 2012, pp. 1-5. high traffic loads, and 2) ensure the deployment of least doi:10.1109/NCC.2012.6176843 amount of components within the network. In terms of [7] N. Sambo, N. Andriolli, A. Giorgetti, I. Cerutti, L. Val- the lightpath setup times, P-OCP outperforms H-OCP at carenghi, P. Castoldi and F. Cugini, “GMPLS-Controlled Dynamic Translucent Optical Networks,” IEEE Network, all traffic loads. However, at high traffic loads, the setup Vol. 23, No. 3, 2009, pp. 34-40. time for H-OCP starts to decrease slightly owing to high doi:10.1109/MNET.2009.4939261 BP, as longer lightpaths are more likely to be blocked; [8] R. Martinez, C. Pinart, F. Cugini, N. Andriolli, L. Val- whereas, for P-OCP, the setup time increases with the carenghi, P. Castoldi, L. Wosinska, J. Comellas and G. traffic load owing to increased number of lightpaths be- Junyent, “Challenges and Requirements for Introducing ing established. Hence, it can be concluded that there Impairment-Awareness into the Management and Control exists a trade-off between obtaining a specific block- Planes of ASON/GMPLS WDM Networks,” IEEE Com- ing performance and lightpath establishment setup time. munications Magazine, Vol. 44, No. 12, 2006, pp. 76-85. doi:10.1109/MCOM.2006.273103 Overall, compared to H-OCP, P-OCP is able to maintain lower setup times and BP at all traffic loads. Finally, the [9] E. Salvadori, Y. Ye, C. V. Saradhi, A. Zanardi, H. Woes- ner, M. Carcagnì, G. Galimberti, G. Martinelli, A. Tanzi proposed OCP is compatible and suitable for controlling and D. La Fauci, “Distributed Optical Control Plane Ar- the operations of the electro-optical hybrid translucent chitectures for Handling Transmission Impairments in node. Transparent Optical Networks,” IEEE Journal of Light- wave Technology, Vol. 27, No. 13, 2009, pp. 2224-2239. 6. Acknowledgements doi:10.1109/JLT.2008.2006066 [10] F. Cugini, N. Sambo, N. Andriolli, A. Giorgetti, L. Val- The authors would like to thank the AICTE for funding carenghi, P. Castoldi, E. Le Rouzic and J. Poirrier, “En- this work under research promotion scheme (RPS-11/ hancing GMPLS Signalling Protocol for Encompassing 2008-09). Quality of Transmission (QoT) in All-Optical Networks,” IEEE Journal of Lightwave Technology, Vol. 26, No. 19, 2008, pp. 3318-3328. doi:10.1109/JLT.2008.925674 REFERENCES [11] N. Sambo, F. Cugini, N. Andriolli, A. Giorgetti, L. Val- [1] C. V. Saradhi, S. Zaks, R. Fedrizzi, A. Zanardi and E. carenghi and P. Castoldi, “Lightweight RSVP-TE Exten- Salvadori, “Practical and Deployment Issues to Be Con- sions to Account for shared Regenerators in Translucent

Optical Networks,” Proceedings of IEEE Photonics in Teletraffic Performance of Wavelength-Routed Optical Switching (PIS), San Francisco, 2007, pp. 35-36. Networks,” IEEE Journal of Lightwave Technology, Vol. doi:10.1109/PS.2007.4300731 17, No. 10, 1999, pp. 1713-1723. doi:10.1109/50.793740 [12] F. Cugini, N. Sambo, N. Andriolli, A. Giorgetti, L. Val- [17] S. P. Singh, S. Kar and V. K. Jain, “Performance of All- carenghi, P. Castoldi, E. L. Rouzic and J. Poirrier, Optical WDM Network in Presence of Four-Wave Mix- “GMPLS Extensions to Encompass Shared Regenerators ing, Optical Amplifier Noise, and Wavelength Converter in Transparent Optical Networks,” Proceedings of IEEE Noise,” Fiber & Integrated Optics, Vol. 26, No. 2, 2007, European Conference and Exhibition of Optical Commu- pp. 79-97. doi:10.1080/01468030601131477 nication (ECOC), Berlin, 16-20 September 2007, pp. 1-2. [18] K. J. Kwak and E. G. Coffman, “Retransmission in OBS [13] L. Wang, J. Zhang, Y. Zhao and W. Gu, “Study of Opti- Networks with Fiber Delay Lines,” Proceedings of IEEE cal Control Plane for Translucent WDM Networks,” International Conference on Broadband Communications, Photonic Network Communications, Vol. 20, No. 1, 2012, Networks and Systems (BROADNETS), Raleigh, 10-14 pp. 64-74. doi:10.1007/s11107-010-0246-2 September 2007, pp. 243-249. [14] D. Li, Y. Lee, and J. Gao, “Evaluation of IGP Extensions doi:10.1109/BROADNETS.2007.4550431 for Wavelength Switching Optical Networks,” IETF [19] MRV Communications, “The Low Latency Network De- Internet Draft, 2007, pp. 1-16. sign Considerations,” White Paper, MRV-AN-Low La- http://tools.ietf.org/pdf/draft-li-ccamp-wson-igp-eval-01.p tency, Optical Communication Systems, 2010. df http://www.mrv.com/library/docs/PDF72/MRV-WP [15] S. Iyer and S. P. Singh, “Impact of Combined Nonlineari- Low-Latency.pdf ties and ASE Noise on Performance of 10 Gbps All Op- [20] J. He, M. Brandt-Pearce and S. Subramaniam, “QoS- tical Star WDM Networks,” Scientific Research, Commu- Aware Wavelength Assignment with BER and Latency nications and Network, Vol. 3, No. 4, 2011, pp. 235-249. Guarantees for Crosstalk Limited Networks,” Proceed- doi:10.4236/cn.2011.34028 ings of IEEE International Conference on Communica- [16] B. Ramamurthy, D. Datta, H. Feng, J. P. Heritage and B. tions (ICC), Glasgow, 24-28 June 2007, pp. 2336-2341. Mukherjee, “Impact of Transmission Impairments on the doi:10.1109/ICC.2007.392